[clang] [llvm] [Clang][RISCV] Add scalar widening zip intrinsics (PR #238758)

via cfe-commits cfe-commits at lists.llvm.org
Wed Nov 4 03:47:51 PST 2026


https://github.com/Michael-Chen-NJU created https://github.com/llvm/llvm-project/pull/238758

Add the scalar Bitmanip intrinsics `__riscv_wzip8p_64` and `__riscv_wzip16p_64` using the existing packed zip intrinsics.


>From 61f643f04d68c31adc32c8c6bb84d2fb185a9535 Mon Sep 17 00:00:00 2001
From: Michael-Chen-NJU <2802328816 at qq.com>
Date: Wed, 4 Nov 2026 19:46:50 +0800
Subject: [PATCH] [Clang][RISCV] Add scalar widening zip intrinsics

---
 clang/lib/Headers/riscv_packed_simd.h         | 11 +++++
 clang/test/CodeGen/RISCV/rvp-intrinsics.c     | 48 +++++++++++++++++++
 .../riscv_packed_simd.c                       | 14 ++++++
 3 files changed, 73 insertions(+)

diff --git a/clang/lib/Headers/riscv_packed_simd.h b/clang/lib/Headers/riscv_packed_simd.h
index 89f678a63e94e6..10c253741587a9 100644
--- a/clang/lib/Headers/riscv_packed_simd.h
+++ b/clang/lib/Headers/riscv_packed_simd.h
@@ -972,6 +972,17 @@ __packed_zip4(pzip_u8x8, uint8x8_t, uint8x4_t)
 __packed_zip2(pzip_i16x4, int16x4_t, int16x2_t)
 __packed_zip2(pzip_u16x4, uint16x4_t, uint16x2_t)
 
+/* Scalar Widening Zip */
+static __inline__ uint64_t __DEFAULT_FN_ATTRS
+__riscv_wzip8p_64(uint32_t __rs1, uint32_t __rs2) {
+  return (uint64_t)__riscv_pzip_u8x8((uint8x4_t)__rs1, (uint8x4_t)__rs2);
+}
+
+static __inline__ uint64_t __DEFAULT_FN_ATTRS
+__riscv_wzip16p_64(uint32_t __rs1, uint32_t __rs2) {
+  return (uint64_t)__riscv_pzip_u16x4((uint16x2_t)__rs1, (uint16x2_t)__rs2);
+}
+
 /* Packed Unzip */
 __packed_unzipe4(punzipe_i8x4, int8x4_t, int8x8_t)
 __packed_unzipo4(punzipo_i8x4, int8x4_t, int8x8_t)
diff --git a/clang/test/CodeGen/RISCV/rvp-intrinsics.c b/clang/test/CodeGen/RISCV/rvp-intrinsics.c
index 956d80beb1228b..0a64157fbd2ca4 100644
--- a/clang/test/CodeGen/RISCV/rvp-intrinsics.c
+++ b/clang/test/CodeGen/RISCV/rvp-intrinsics.c
@@ -204,6 +204,54 @@ uint32_t test_asubu_u32(uint32_t a, uint32_t b) {
   return __riscv_asubu_u32(a, b);
 }
 
+/* Scalar Widening Zip */
+
+// CHECK-LABEL: define dso_local i64 @test_wzip8p_64(
+// RV32-SAME: i32 noundef [[RS1:%.*]], i32 noundef [[RS2:%.*]]) #[[ATTR0]] {
+// RV64-SAME: i32 noundef signext [[RS1:%.*]], i32 noundef signext [[RS2:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32 [[RS1]] to <4 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32 [[RS2]] to <4 x i8>
+// CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i8> [[TMP0]], <4 x i8> [[TMP1]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[SHUFFLE_I]] to i64
+// CHECK-NEXT:    ret i64 [[TMP2]]
+//
+uint64_t test_wzip8p_64(uint32_t rs1, uint32_t rs2) {
+  return __riscv_wzip8p_64(rs1, rs2);
+}
+
+// CHECK-LABEL: define dso_local i64 @test_wzip16p_64(
+// RV32-SAME: i32 noundef [[RS1:%.*]], i32 noundef [[RS2:%.*]]) #[[ATTR0]] {
+// RV64-SAME: i32 noundef signext [[RS1:%.*]], i32 noundef signext [[RS2:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32 [[RS1]] to <2 x i16>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32 [[RS2]] to <2 x i16>
+// CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <2 x i16> [[TMP0]], <2 x i16> [[TMP1]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to i64
+// CHECK-NEXT:    ret i64 [[TMP2]]
+//
+uint64_t test_wzip16p_64(uint32_t rs1, uint32_t rs2) {
+  return __riscv_wzip16p_64(rs1, rs2);
+}
+
+// CHECK-LABEL: define dso_local i64 @test_wzip8p_64_constant(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    ret i64 -105873736745117663
+//
+uint64_t test_wzip8p_64_constant(void) {
+  return __riscv_wzip8p_64(0x87654321, 0xfedcba98);
+}
+
+// CHECK-LABEL: define dso_local i64 @test_wzip16p_64_constant(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    ret i64 -82041822207524063
+//
+uint64_t test_wzip16p_64_constant(void) {
+  return __riscv_wzip16p_64(0x87654321, 0xfedcba98);
+}
+
 /* Scalar Saturating Addition and Subtraction */
 
 // RV32-LABEL: define dso_local i32 @test_sadd_i32(
diff --git a/cross-project-tests/intrinsic-header-tests/riscv_packed_simd.c b/cross-project-tests/intrinsic-header-tests/riscv_packed_simd.c
index d1c002629d006f..24b640756a5f9b 100644
--- a/cross-project-tests/intrinsic-header-tests/riscv_packed_simd.c
+++ b/cross-project-tests/intrinsic-header-tests/riscv_packed_simd.c
@@ -24,6 +24,20 @@ unsigned test_cls_32(int32_t a) { return __riscv_cls_32(a); }
 // RV64-NEXT:   srai
 uint32_t test_rev_32(uint32_t a) { return __riscv_rev_32(a); }
 
+// CHECK-LABEL: test_wzip8p_64:
+// RV32:        wzip8p{{[[:space:]]}}
+// RV64:        zip8p{{[[:space:]]}}
+uint64_t test_wzip8p_64(uint32_t rs1, uint32_t rs2) {
+  return __riscv_wzip8p_64(rs1, rs2);
+}
+
+// CHECK-LABEL: test_wzip16p_64:
+// RV32:        wzip16p{{[[:space:]]}}
+// RV64:        zip16p{{[[:space:]]}}
+uint64_t test_wzip16p_64(uint32_t rs1, uint32_t rs2) {
+  return __riscv_wzip16p_64(rs1, rs2);
+}
+
 // CHECK-LABEL: test_slx_32:
 // RV32:        slx{{[[:space:]]}}
 // RV64:        andi



More information about the cfe-commits mailing list